Background modeling based on YCbCr color space and gesture shadow elimination
-
摘要: 为了提高动态手势检测的精确度,本文将基于YCbCr颜色空间的混合高斯背景建模应用于动态手势识别中,并且提出手势阴影消除的有效算法。首先,对待检测视频帧通过抠图抠出手势图像,在YCb'Cr'颜色空间进行椭圆拟合,统计建立椭圆肤色模型,继而在YCbCr颜色空间进行混合高斯背景建模检测出动态手势,点乘原图像得到含有阴影的RGB手势图像,对检测出的含有阴影的手势图像利用已建立的椭圆肤色模型进行阴影消除,最后将手势图像连成视频序列。实验结果表明,该算法在复杂背景下进行动态手势的检测率可达91.4%,高出传统方法10%左右,能够满足动态手势检测基本要求,且具有较高的实用价值。Abstract: To improve the accuracy of the dynamic gesture detection, Gaussian mixture background modeling based on YCbCr color space is applied to the dynamic gesture recognition, and the effective gesture shadow elimination algorithm is proposed. First of all, the gesture image is cut out from video frame to be detected, and space ellipse fitting is developed in YCb'Cr' color. Oval color model is established statistically, and then dynamic gesture in the YCbCr color space through Gaussian mixture background modeling is detected. Original image is dotted product to get the gesture RGB image containing shadows. The shadows contained in the detected gestures image were eliminated by using ellopse color model, and finally we take gesture images together into a video sequence. Experimental results show that in the algorithm of dynamic gesture detection rate is 91.4% under a complex background, about 10% higher than that by the traditional methods. So it can meet the basic requirements of dynamic gesture detection, and has a high practical value.
-
图 2 肤色在Cb′-Cr′坐标系下的投影
Figure 2. Projection of oval color in color Cb′-Cr ′coordinate system
表 1 检测结果统计
Table 1. Statistics of test results
TP FP DR 传统高斯模型检测 57 13 81.4% 本文算法 64 6 91.4% 
下载: 导出CSV
-
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] -
下载:
点击查看大图
图(7) / 表(1)
计量
- 文章访问数: 1670
- HTML全文浏览量: 295
- PDF下载量: 738
- 被引次数: 0
